Divergence Between Absolute Emissions Benchmarks and Output-Adjusted Assessments: Evidence from Korea

Sustainability-linked loans and bonds reward borrowers meeting emissions targets, mostly set in absolute terms, and Korea is establishing verification for green-transition finance. Yet an unindexed absolute benchmark cannot distinguish reducing emissions per unit of output from contracting production. Using site-level physical output from Korea's mandatory environmental disclosure system, we observe both dimensions for energy-intensive manufacturers and power generators. Among 513 firm-year pairs attaining a hypothetical 4.2% annual absolute benchmark, 69.4% fail an intensity benchmark at the same rate, while an elasticity-adjusted assessment yields a model-dependent output-neutral shortfall rate of 47.2%. The intensity shortfall rate stays at 68.8% without power generation, and a broader revenue sample gives 50.0%, both above placebo thresholds. An exact identity decomposes emissions changes into physical output, energy intensity and emissions per unit of energy. Replacing physical output with deflated revenue understates the output-driven portfolio share by 11.5 points (intensity reference). Where physical output is undisclosed, a proxy rule combining energy use and real revenue screens attainments better than accepting all attainments and, in Korea, shows no detectable portfolio-share bias, though it matches revenue alone on individual output-neutral verdicts and does not explain firm-level residual reductions. These findings describe a selected Korean sample under a hypothetical benchmark.

Version 7.7.1 (21 October 2026). Revised version: new title, physical-output sample with the intensity shortfall rate as the main measure, multi-year and boundary robustness, and an R replication package (https://doi.org/10.5281/zenodo.23501951). Supersedes the version of 2 October 2026.